OICC Press International Journal of Mathematical Modelling & Computations 2228-6233 16 2 2026 06 30 Deep Reinforcement Learning-based Fusion of GRU and NCF for personalized Recommendations 10.57647/ijm2c.2026.1602.11 EN Mahshid Salehi Department of Computer Engineering, Ma.C, Islamic Azad University, Mashhad, Iran Hassan Shakeri Department of Computer Engineering, Ma.C, Islamic Azad University, Mashhad, Iran https://orcid.org/0000-0002-0355-3386 Seyyed Javad Seyyed Mahdavi Chabok Department of Electrical Engineering, Ma.C, Islamic Azad University, Mashhad, Iran Journal Article 2026 06 30 Recommender systems are widely used to deliver efficient, customized content and services, making them effective solutions for managing information overload. Over the past decade, e-commerce apps, including movie apps, have researched and implemented many recommendation algorithms. However, data sparsity is a significant challenge for many movie recommender datasets. This paper introduces a hybrid recommender system that combines temporal data, user clustering, and item features with user-item ratings to enhance recommendation evaluation metrics while reducing the effects of data sparsity. The Gated Recurrent Unit (GRU) model uses user-item ratings and timestamps to predict temporal behavior patterns. NCF makes the second initial prediction using user-item ratings and user clusters. In the third initial prediction, NCF uses movie genres and user-item ratings. The final recommendation score is the weighted average of three model outputs. To optimize the fusion of these outputs for improved prediction, four deep reinforcement learning algorithms, deep Q-learning, Deep Deterministic Policy Gradient (DDPG), Twin Delayed Deep Deterministic Policy Gradient (TD3), and Proximal Policy Optimization (PPO), are employed. These agents dynamically learn the weights for the fusion process. The experiments utilize MovieLens 100K and 1M datasets. The proposed approach's recommendation quality improvement is measured by precision, recall, MAE, and RMSE.